[Guide] Improving Battery Life on Windows [+Enabling Deeper C States]

Discussion in 'Windows OS and Software' started by Che0063, Apr 14, 2018.

?

What was your increase in battery life after following this guide?

  1. None (decreased)

    16.2%
  2. 20%

    13.5%
  3. 40%

    29.7%
  4. 50% +

    40.5%
  1. Emtee_

    Emtee_ Notebook Enthusiast

    Reputations:
    7
    Messages:
    39
    Likes Received:
    19
    Trophy Points:
    16
    I don't know about that, my laptop idles prettty quickly with less than <0.25% cpu cycles after like 30 secs to a minute after resuming from Standby. I don't have that kind of runaway usage here or whatever its causing it.

    I can do whatever, but unless I start screwing around with thunderbolt device in device manager I'll never ever recover back into C10, until I actually hibernate again and resume.

    If the dGPU actually causes issues that probably indicates that PCIe ASPM is not working properly. It should work fine in general, also for NVMe but they also turn themselves down.

    No clue about modern standby, my laptop doesn't support that :p but yeah 1.1% an hour is horrible, I can hibernate my razer and comes back out 100% after like 18 hours of not having it touched hehe. Unless you need it, better disable it imho.
     
    Last edited: Jul 4, 2020
  2. maffle

    maffle Notebook Evangelist

    Reputations:
    160
    Messages:
    518
    Likes Received:
    465
    Trophy Points:
    76
    @Emtee_ and with what did you measure the "<0.25% cpu cycles" you are speaking about? Like I said above, the issue is not shown in task manager or for example process explorer. Both would show 99% idle after S3 wake up. Only TS would hint at this runaway issue, showing a c0% activity on one random core, most core0, which other task tools wont show. I dont know why this is the case, but I assume this hints that this is a process running in the firmware/EC or in a driver (mostly the usb driver or TB), which wont show in process managers. ASPM is deactivated on most laptops and tablets, so that is kinda redundant. I dont accept using hibernate, that is a total no-go and just nonsense for a laptop. Hibernate wake up takes about 15 seconds or so on the 9570 with 16GB of RAM.
     
  3. Emtee_

    Emtee_ Notebook Enthusiast

    Reputations:
    7
    Messages:
    39
    Likes Received:
    19
    Trophy Points:
    16
    I see, my hibernate literally takes 2~3 seconds (Ok, not much running the background atm but still acceptable)

    ASPM disabled? Mhh.. Isn't that basically PCIe Power Saving in windows power plan setting? As that has a definite effect here in terms of heat production.

    I just looked now and S3'ed again, all my cores are instantly below <1.0 C0% except core0 which remains a little above 1.0 sometimes.
     
  4. maffle

    maffle Notebook Evangelist

    Reputations:
    160
    Messages:
    518
    Likes Received:
    465
    Trophy Points:
    76
    That sounds not right and more like S3 wake time. 2-3 seconds hibernate wake time? I highly doubt that. Hibernate wake up would mean a cold boot, so it is not just the entire boot cycle, it is also load of the RAM from the SSD. That cant take 2-3 seconds. S3 wake up is already 2-3 seconds.

    So this c0% activity after S3 sleep may just be a specific bug on Dell laptops, or just the 9570. I tested it on 3 other 9570 and they all showed the same issue. The 12% c0% activity then obviously would lower the c states and result in around 1.5W drain. Like I said, deactivating all USB devices cleared this broken state sometimes. So it may be the same issue, but manifesting differently on yours.

    No it is not. There are 4 power states of PCIe also link speed.
     
  5. Emtee_

    Emtee_ Notebook Enthusiast

    Reputations:
    7
    Messages:
    39
    Likes Received:
    19
    Trophy Points:
    16
    Well my C10 states tell me the contrary after resume from hibernation, so does my laptops led indicator which stays lit up (breathing) in regular S3 standby, whilst hibernated the unit is fully off, and still have 99 to 100% bat remaining after more almost a full day of not having it used, versus S3 standby where I can clearly see a few % have been depleated.

    Also S3 resume is merely a second if not shorter, what else is it taking 3~ seconds to resume then? S3 doesn't show the circular loading icon either which only occurs on a cold boot or resume from hibernation.

    Last but not least, it takes a tad longer to go to sleep as you can clearly see it has a delay writing down the contents to the disk.

    If a cold boot barely takes 6~ seconds (measured right after post to login screen appearing), why is it so hard to believe that suspend from disk only takes a few seconds tops? Its quite possible these days.

    Remember this is a brand new gaming grade notebook, and the POST on this laptop is incredibly fast as well, next to my pretty optimized windows 10 install.

    Edit: Ok I remeasured from literally pushing the button to login screen, its roughly 4~ seconds including full POST ;-)
    It it acceptable for me? Yea sort of, is it acceptable in general? No. Someone did a sloppy job no doubt 1 second versus 4~5 seconds is still a huge difference in retrospect. Battery wise I doubt it makes much of a difference, but it does depend on the interval on non-usage.

    The situation on my desktop is this: 4~5 seconds post, for various (stability) reasons and otherwise I dont want to reduce this (although I can), and resume from hibernation right after POST is even faster than my new notebook, looking at 2~3 seconds tops, but this system is also way faster.

    Pretty sure last time one of my external PCIe cards did not play well with PCIe ASPM, I fiddled around with the bios options (Changing ASPM to Native, or bios controlled) and in which case when it was on native I could disable ASPM through the OS in my power plan through this very option. Not sure, this is just my observation.

    It is often more of a nuisance and totally not worth it on a desktop but on a laptop I can not really understand why it would not support ASPM, even my Intel NUC does and that is theoretically just a very small form factor PC.
     
    Last edited: Jul 4, 2020
  6. maffle

    maffle Notebook Evangelist

    Reputations:
    160
    Messages:
    518
    Likes Received:
    465
    Trophy Points:
    76
    Because waking up from hibernate is cold boot / POST PLUS loading the hibernate file to ram. So it cant be shorter than the 6 seconds, 6 plus 6 seconds sound about right, resulting in 12-15 seconds. I just did a hibernate wake test with 7GB of RAM usage on the 9570. It took 23 seconds to wake up from hibernate to login screen. After 6 seconds the pre POST was done, after 12 seconds Windows was beginning to boot, took then another 10 seconds to load the hibternate file. Mostly another reason I love Dell laptops, not.
     
  7. Emtee_

    Emtee_ Notebook Enthusiast

    Reputations:
    7
    Messages:
    39
    Likes Received:
    19
    Trophy Points:
    16
    Bleh, that's long. Sorry to hear I guess hibernation is no solution for you due to insanely long total loading time.

    Doesn't look like that laptop is that cheap either.

    Just curious what the sustained speed on your disk is? My laptop and desktop both around 3.6GB/s sustained (luckily hibernate file is barely to not fragmented if pre-allocated 100%)

    On regular usage I rarely go beyond 10GB mem usage anyhow.
     
  8. maffle

    maffle Notebook Evangelist

    Reputations:
    160
    Messages:
    518
    Likes Received:
    465
    Trophy Points:
    76
    @Emtee_ I use a Samsung 970 Evo in that laptop.
     
  9. IntelUser

    IntelUser Notebook Deity

    Reputations:
    362
    Messages:
    1,639
    Likes Received:
    67
    Trophy Points:
    66
    Emtee, if your laptop wakes up nearly instantly after opening lid, then it supports modern standby. Ever since 4th Gen U core, Intel chips support what's called a S0iX state, which is a combination that tries to bring the benefits of S3 like low power and S0 like wake up time. The support is not perfect though. I think the reason is Intel chips still have a way to go in lowering S0iX power.

    Supporting C10 is a key part of that, but as @maffle device indicates the U series uses a lot less power in all states than H series chips do. Also, the U chips have an on-package PCH while H chips need a separate one.
     
  10. maffle

    maffle Notebook Evangelist

    Reputations:
    160
    Messages:
    518
    Likes Received:
    465
    Trophy Points:
    76
    I just randomly now again had this package "c state bug" with my XPS 9570 yesterday, and I have no idea, whats triggering it.

    [​IMG]

    All was working normally for days, lots of hibernate wake ups in between, and *poof* yesterday just that. I noticed it actually because of fans always running, looked into TS, and saw the permanent 2.5W usage and all package c states off... even with 0.4% c0% activity.

    No USB device on the laptop, no TB. Had to reboot or shutdown to make it work again.

    Before reboot, I tried to disable and enable again several devices in device manager, like the Nvidia GPU, wifi, bluetooth, with no effect.

    I am using the latest Intel SATA AHCI RST driver right now, version 17.9.1, may it be a try worth using the default MS SATA AHCI driver? The Intel AHCI driver isnt really necessary at all, right? Any disadvantage not using it?
     
    Last edited: Jul 28, 2020

Share This Page